fix(bedrock): Handle streaming tool calling with no input arguments

- Fix isEmpty() check in ConverseApiUtils to properly handle null partialJson
- Add workaround for streaming tool calls with empty arguments by defaulting to {}
- Add parameterized tests for multiple Bedrock models (Nova Pro and Claude Sonnet)
- Update AWS SDK and Bedrock runtime versions to 2.31.65

Fixes #1878

Signed-off-by: Christian Tzolov <[email protected]>

Author: tzolov

models/spring-ai-bedrock-converse/src/main/java/org/springframework/ai/bedrock/converse/api/ConverseApiUtils.java

@@ -421,8 +421,7 @@ public List<ToolUseEntry> toolUseEntries() {
 		}
 
 		public boolean isEmpty() {
-			return (this.index == null || this.id == null || this.name == null
-					|| !StringUtils.hasText(this.partialJson));
+			return (this.index == null || this.id == null || this.name == null || this.partialJson == null);
 		}
 
 		ToolUseAggregationEvent withIndex(Integer index) {
@@ -451,7 +450,9 @@ ToolUseAggregationEvent appendPartialJson(String partialJson) {
 		}
 
 		void squashIntoContentBlock() {
-			this.toolUseEntries.add(new ToolUseEntry(this.index, this.id, this.name, this.partialJson, this.usage));
+			// Workaround to handle streaming tool calling with no input arguments.
+			String json = StringUtils.hasText(this.partialJson) ? this.partialJson : "{}";
+			this.toolUseEntries.add(new ToolUseEntry(this.index, this.id, this.name, json, this.usage));
 			this.index = null;
 			this.id = null;
 			this.name = null;

models/spring-ai-bedrock-converse/src/test/java/org/springframework/ai/bedrock/converse/client/BedrockNovaChatClientIT.java

@@ -23,6 +23,8 @@
 import java.util.stream.Collectors;
 
 import org.junit.jupiter.api.Test;
+import org.junit.jupiter.params.ParameterizedTest;
+import org.junit.jupiter.params.provider.ValueSource;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 import reactor.core.publisher.Flux;
@@ -186,12 +188,15 @@ void toolAnnotationWeatherForecast() {
 		assertThat(response).contains("20 degrees");
 	}
 
-	@Test
-	void toolAnnotationWeatherForecastStreaming() {
+	// https://github.com/spring-projects/spring-ai/issues/1878
+	@ParameterizedTest
+	@ValueSource(strings = { "amazon.nova-pro-v1:0", "us.anthropic.claude-3-7-sonnet-20250219-v1:0" })
+	void toolAnnotationWeatherForecastStreaming(String modelName) {
 
 		ChatClient chatClient = ChatClient.builder(this.chatModel).build();
 
 		Flux<ChatResponse> responses = chatClient.prompt()
+			.options(ToolCallingChatOptions.builder().model(modelName).build())
 			.tools(new DummyWeatherForecastTools())
 			.user("Get current weather in Amsterdam")
 			.stream()
@@ -257,6 +262,7 @@ public static class Config {
 		public BedrockProxyChatModel bedrockConverseChatModel() {
 
 			String modelId = "amazon.nova-pro-v1:0";
+			// String modelId = "us.anthropic.claude-3-7-sonnet-20250219-v1:0";
 
 			return BedrockProxyChatModel.builder()
 				.credentialsProvider(EnvironmentVariableCredentialsProvider.create())

pom.xml

@@ -261,8 +261,8 @@
 		<kotlin.version>1.9.25</kotlin.version>
 
 		<!-- NOTE: keep bedrockruntime and awssdk versions aligned -->
-		<bedrockruntime.version>2.31.26</bedrockruntime.version>
-		<awssdk.version>2.29.29</awssdk.version>
+		<bedrockruntime.version>2.31.65</bedrockruntime.version>
+		<awssdk.version>2.31.65</awssdk.version>
 
 		<djl.version>0.32.0</djl.version>
 		<onnxruntime.version>1.19.2</onnxruntime.version>